In a Lowland Bornean Rain Forest: Fig Wasps Are the High Fliers1

BIOTROPICA 32(3): 515–522 2000

The Flight Heights of Chalcid Wasps (Hymenoptera, Chalcidoidea) in a Lowland Bornean Rain Forest: Fig Wasps are the High Fliers1

Stephen G. Compton Centre for Biodiversity and Conservation, School of Biology, The University of Leeds, Leeds LS2 9JT, England Martin D. F. Ellwood University Museum of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, England; and Danum Valley Field Centre, Sabah, Malaysia Andrew J. Davis Department of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, England; and Danum Valley Field Centre, Sabah, Malaysia and Kathleen Welch Centre for Biodiversity and Conservation, School of Biology, The University of Leeds, Leeds LS2 9JT, England

ABSTRACT Tropical rain forests are characterized by their rich plant diversity and highly diverse insect faunas containing mainly rare species. Phytophagous and parasitoid insects utilizing such fragmented resources often must travel considerable distances to find suitable hosts. For small, weak-flying insects, entry into the fast-flowing air above the canopy can provide one way by which long-distance dispersal is achieved. Using sticky traps placed at different heights in a lowland rain forest of Borneo, we compared the diurnal and nocturnal flight heights of chalcids, a group of mainly very small parasitoids and phytophages, to determine if the air above the canopy was used for dispersal. Most families were represented throughout the range of trap heights, including those above the general canopy. A higher proportion of individuals were trapped above the canopy at night than during the day. Fig wasps were exceptional in that they were trapped almost entirely above the canopy. They included species associated with host trees that do not fruit in the canopy, suggesting that these short-lived, slow-flying insects actively fly up above the canopy and then use the wind to passively carry them the long distances needed to reach their highly localized and ephemeral hosts. Once the fig wasps detect the species-specific volatiles released by their host figs, they then may fly down into the canopy, where the lower wind speeds would allow them to fly actively upwind to their hosts.

Key words: Agaonidae; dispersal; Ficus; fig wasp; flight height; Mymaridae; parasitoid; rain forest; Sabah; vertical stratification.

ADULT FEMALE FIG WASPS (Hymenoptera, Chalci- or requiring their services. The difficulties faced by doidea, Agaonidae) are the sole pollinators of fig the wasps are made more urgent by their short trees (Ficus spp., Moraceae). Each species of fig adult life spans (probably 48 hours at most) and tree, of which there are Ͼ 700, usually is pollinated the typically low densities of rain forest fig trees, by a single species of fig wasp that is not associated particularly strangler figs (Mawdsley et al. 1998); with any other fig tree species (Janzen 1979, Wie- it is clear that rain forest fig wasp pollinators often bes 1979). Fig wasps are able to successfully track need to both cover large distances quickly and ac- a resource that is highly ephemeral and dispersed, curately to home-in on those hosts having figs at especially in rain forests, where many fig trees pro- the correct stage of development. The same prob- duce large, highly synchronized crops attracting lems are faced by the numerous fig wasps that do thousands of pollinating wasps over periods of only not pollinate figs (currently also placed in Agaon- a few days (Nason 1996). Fruiting phenologies vary idae), which are either seed predators or parasitoids, among fig tree species, but at the population level, although the majority of these species probably fruiting is typically asynchronous, and continues have longer adult life spans. throughout the year. Consequently, at any one When they are ready to be pollinated, fig trees time, only a small proportion of the trees belonging release species-specific volatiles that attract their to any one species either are releasing pollinators particular pollinators, and no others (Ware et al. 1993, Ware & Compton 1994a). Because they are 1 Received 19 January 1999; revision accepted 28 May so small, fig wasps have low flight speeds, and 1999. therefore have no control over their direction of

515 516 Compton, Ellwood, Davis, and Welch

flight once they are in air moving greater than a sions. Such movements, at least when the figs are few meters per second (Ware & Compton 1994b). produced within and below the canopy, will be Given the often great distances between fig trees, largely limited in distance by the flight speeds of there may be a trade-off between needing faster- the wasps, which nonetheless can control the di- moving air to aid random long-distance dispersal rection of their flights. Alternatively, fig wasps may and slower wind speeds in which the wasps can take advantage of the faster-moving air that is avail- control their direction of flight toward suitable host able above the canopy to facilitate long-distance trees. Studies in South Africa and India (Ware & movement, initially flying upwards but then de- Compton 1994b,c; P. Kathuria & S. G. Compton, scending after a certain time period or once they pers. obs.) have suggested that in savanna environ- have detected attractive volatiles. For attractants ments, fig wasps disperse upwards and downwind from figs produced in or below the canopy to reach after emergence and are carried by the wind until the general air column above the forest, it is nec- they detect odors of fig trees with fruits at the cor- essary for some upward movement of air to occur; rect stage of development. They then descend out however, this may be quite common (Whitmore of the air column, and by moving in and above 1984, Kira & Yoda 1989). Movement above the the ground flora (where wind speeds are lower; canopy by fig wasps would allow greater areas of Whitmore 1984, Sutton 1989), they make their the forest to be covered, but poses problems for way upwind to the source. such delicate insects, especially during the day Similar movement patterns are unlikely to ap- when temperatures can be much higher, and hu- ply in a rain forest environment. Here, figs are pro- midities lower, than beneath the canopy. Flight duced at a great range of heights, from below the above the canopy therefore may be an option only ground to emergents rising above the general can- for night-flying fig wasps and other chalcids of a opy at 50 meters or more. Wind speeds are typi- similar size. cally lower than in savanna situations, while at the There are few studies of the heights at which lower levels of the forest, the air is often effectively fig wasps and other chalcids fly. In open situations still. Understory flight by fig wasps therefore could among savanna-type vegetation, approximately be more easily controlled and directed, but the dis- even densities of flying fig wasps were found within tances they could travel would be limited by their the air column from the ground to four meters low flight speed (recorded as Ͻ 37 cm/sec; Comp- (Ware & Compton 1994a). Using light traps in ton 1993) and short life span. This maximum rain forest of Zaire and Sabah, Sutton and others flight speed translates to Ͼ 1 km/h, if flight could (Sutton & Hudson 1980; S. L. Sutton, pers. be maintained continuously, and potentially would comm.) found that night-flying fig wasps were allow them to travel several kilometers during their more abundant within the canopy than at ground one or two days of adult life. The fig wasps, how- level, a conclusion also reached by Kato et al. ever, clearly cannot orient to suitable hosts over (1995). In contrast, Ng (1978), based on malaise such long distances and the chances of coming trap catches in western Malaysia, concluded that within the attractant radius of a tree with figs at hymenopterans, of which chalcids were the most the correct stage will be correlated strongly with abundant group, mainly flew close to the ground the distances that can be traveled (upwind directed and during the day. movements toward figs of 100 m have been re- Although night-flying fig wasps and small corded, but the true extent of the attractant radii numbers of other chalcids are readily attracted to is unknown; P. Kathuria & S. G. Compton, pers. ultraviolet lights positioned from ground level to obs.). The same problems arise for other small, above the general canopy in rain forests (Kato et weak-flying insects in rain forests, such as other al. 1995; S. G. Compton & M. D. F. Ellwood, chalcid hymenopterans. Many of these species, pers. obs.), the large potential radius of attraction however, are likely to be associated with resources produced by the lights makes this trapping method that are not as spatially concentrated as figs, and unsuitable for determining the heights at which so they are unlikely to be as dependent on long- wasps are moving before they are attracted. Fur- distance flights to reach potential hosts. thermore, most chalcids, including some fig wasps, Given the difficulties faced by rain forest fig are day-flying, and so cannot be collected in this wasps, how do they manage to get from tree to way. In this study, clear plastic sticky traps posi- tree? Dispersing fig wasps may stay more or less at tioned at various heights in the canopy were used the heights where they are released by the trees, to examine the following hypotheses relating to the essentially searching the forest in only two dimen- dispersal of fig wasps and other chalcids in rain Rain Forest Chalcid Flight Heights 517 forests: (1) fig wasps are more likely than other n-heptanal to remove the adhesive (Henshaw chalcids to travel above the canopy, because the 1993), then stored in 70 percent ethanol. The traps resources they need to locate are so localized that were examined twice daily between 26 June and 5 long-distance dispersal is required; (2) night-flying July, giving nine daytime and nine nighttime trap- fig wasps and other chalcids are more likely to fly ping periods. Continuous trapping was extended above the canopy than day-flying species because for an additional ten days (period 2). At the end of the extreme physical conditions they would en- of this period, the acetate sheets were removed and counter above the canopy during the day; and (3) covered by thin transparent plastic sheets for trans- fig wasps disperse between trees by flying above the portation back to the laboratory and subsequent canopy, irrespective of the heights of their host figs. identification. Placement into families of Chalci- doidea followed Boucek (1988), with all fig wasps STUDY SITE AND METHODS relegated to Agaonidae and pollinating species placed in the subfamily Agaoninae. Trapping was carried out in Borneo at the Danum Valley Field Centre in the Malaysian state of Sabah RESULTS in June and July 1998. The traps were suspended from a 40-m observation platform erected in low- Representatives of eleven families of Chalcidoidea land riverine dipterocarp forest adjacent to the field were present on the traps, with the better-repre- center. In this area, the canopy surface is irregular, sented families generally found throughout the with numerous emergents (Newbery et al. 1992). range of trap heights (Table 1). The traps situated Five cylindrical traps (diameter 14 cm, length 90 in the overstory above the general canopy (traps 4 cm), each comprising six-A4 clear acetate sheets at- and 5) and at ground level (trap 1) captured more tached to a wire mesh frame, were covered with chalcids than those in the canopy (Table 1). Num- Oecotac௡ insect trapping adhesive. The adhesive bers trapped are not straightforward indicators of and underlying plastic were shiny, rather than en- densities, however, because the more rapid air tirely transparent, which may have influenced the movement above the general canopy would have behavior of some flying insects. They were sus- resulted in more air being ‘‘sampled’’ by the traps pended at 9-m intervals vertically above each other there. from ground level to 37 meters above the ground. When fig wasps were distinguished from other The main canopy at this location reached ca 30 m, chalcids, very different height profiles were evident. with individual trees ascending above this to ca 50 About as many non-fig wasps were trapped at m. Consequently, traps 4 and 5 were in the over- ground level as on the highest of the traps, despite story (above the general canopy, but below the the lower volumes of air sampled at ground level highest emergents), while 3 and 2 were within and (Fig. 1A). Pollinator fig wasps, in contrast, were slightly below the main canopy layer, respectively. restricted almost entirely to the traps above the Ground-level trap 1 was standing in a small clear- general canopy (Fig. 1B). The small number of ing. Each trap was equipped with a rain guard im- non-pollinator fig wasps that were trapped showed mediately above it. The traps could be lowered for a similar pattern (Table 1). Differences in the pro- examination at ground level when required using a file of abundance with height also were present pulley system. within different families of non-fig wasps (Table 1); During period 1 of the study, the traps were encyrtids (the best represented family), for exam- examined twice daily at approximately dusk and ple, were much more active above the canopy than dawn to differentiate between night- and day-flying mymarids, which were present mainly at ground species. Insects were removed by eye from the traps level (Figs. 1c, d). starting at 0615 and 1625 h, rather than true dusk Relatively fewer mymarids, aphelinids, and tri- and dawn, because light was needed to view the chogrammatids were recorded during period 1, trapped insects. Dawn was at ca 0600 and dusk at when the chalcids were removed directly from the 1815 h. The ‘‘nighttime’’ period therefore included traps at the beginning and end of each day, com- any catches of crepuscular species. When being ex- pared to period 2, when the traps were left in place amined, the traps were suspended at approximately and then removed to the laboratory (Tables 2 and head height and all the small insects present were 3). These three families contained the smallest chal- removed with a fine metal wire prior to subsequent cids (often Ͻ 1 mm in length), and some clearly microscopic examination. Fig wasps and other were being missed during the twice-daily scanning specimens requiring further study were treated in of the traps. 518 Compton, Ellwood, Davis, and Welch 2 5 2–5 4–5 1–5 1–5 1–5 1–5 1–5 1–5 2–5 1–5 1–5 Trap (range) positions llinators belonging ) ¯ x ( 4.7 4.8 2.7 4.0 2.0 3.6 2.9 4.1 2.0 3.4 5.0 3.3 2.0 Trap position 7 1 1 48 74 31 98 26 35 41 232 133 199

FIGURE 1. Captures on sticky traps positioned from ground level (trap 1) to 37 m (Trap 5) in rain forest at Danum Valley Field Centre, Sabah. (A) Chalcids other than ﬁg wasps (Agaonidae); (B) Pollinator ﬁg wasps

5 4 0 7 1 5 (Agaonidae, Agaoninae); (C) Encyrtidae; and (D) My- 93 43 49 13 14 56 25 maridae.

During period 1, when day and nighttime sam- ples were collected separately, the total number of 1 0 8 0 4 1 6 12 49 23 24 16 17 chalcids trapped during the day was almost twice as many as at night, despite the shorter daytime trapping interval (Tables 2 and 3). Therefore, general activity was probably greater during daylight hours, although greater wind speeds during the day 9 1 0 4 0 0 4 3 0 0 could have biased the results, at least for traps 32 11 32 Traps above the canopy, and the traps would be more visible during the day. Fig wasps again provided an exception to this activity pattern, with more individuals being trapped during the night. The heights at which chalcids in general were ﬂying showed 9 0 1 9 0 5 1 0

30 11 41 11 12 some diel variation, with 63 percent of the chalcids on overstory traps 4 and 5 during the day, compared to 82 percent at night. The proportion of the total chalcid captures on the overstory traps (4 and 5 combined) at night was signiﬁcantly higher ␹2 than during the day (Tables 2 and 3; Fig. 2; [1] 1 0 5 0 0 8 0 0 1 2 3ϭ 4Ͻ 5 Total

13 28 52 97 20 46.9, P 0.001). This pattern was still evident when the predominantly nighttime and high-flying ␹2 ϭ Ͻ fig wasps were excluded ( [1] 27.2, P 0.001). Almost all the fig wasps (both pollinators and non- pollinators) were collected above the canopy on traps 4 and 5. Furthermore, the highest trap captured 89 percent of the pollinators at night com- to the subfamilies Epichrysomallinae, Sycoryctinae, and Sycophaginae. ‘‘Others’’ were damaged chalcids that could not be placed to family. Captures of chalcids on sticky traps positioned from ground level (trap 1) to 37 m (trap 5). Periods 1 and 2 were combined. ‘‘Other fig wasps’’ were non-po pared to 29 percent during the day (day:night captures on trap 5 compared to traps 1–4 combined; ␹2 ϭ Ͻ [1] 27.6, P 0.001), suggesting that they tended to fly even higher at night. Aphelinidae Chalcididae Elasmidae Encyrtidae Eulophidae Eupelmidae Mymaridae Pteromalidae Torymidae Trichogrammatidae Others Total chalcids 224 130 96 161 315 926 3.2 1–5 TABLE 1. Fig pollinators Other fig wasps The flight heights of the fig wasps collected on Rain Forest Chalcid Flight Heights 519

TABLE 2. Nighttime captures of chalcids on sticky traps positioned from ground level (trap 1) to 37 m (trap 5). Period 1 only: no Elasmidae, Torymidae, or Trichogrammatidae were recorded during this period. ‘‘Other ﬁg wasps’’ were non-pollinators belonging to the subfamilies Epichrysomallinae, Sycoryctinae, and Sycophaginae.

Traps Trap Trap position positions 12345Total (x¯) (range) Fig pollinators 0 0 0 2 17 19 4.9 4–5 Other fig wasps 0 0 0 0 2 2 5.0 5 Aphelinidae 0 0 0 5 0 5 4.0 4 Chalcididae 0 0 0 1 0 1 4.0 4 Encyrtidae 4 2 3 12 22 43 4.0 1–5 Eulophidae 0 0 0 4 7 11 4.6 4–5 Eupelmidae 0 2 1 1 4 8 3.9 2–5 Mymaridae 0 0 0 1 0 1 4.0 4 Pteromalidae 0 7 0 7 1 15 3.8 2–5 Total chalcids 4 11 4 33 53 105 4.1 1–5 the traps could not be related directly to the heights canopy as those fig wasps that pollinate strangler of the figs they pollinate, because the host relation- figs. ships for many of the species are unknown. Indirect Pollinating fig wasps were represented by at comparisons are nonetheless possible, using the least 21 morphologically distinct taxa on the sticky breeding systems of their host plants. These are traps (based on 68 individuals in suitable condition readily distinguishable, because those wasps that for assignment to taxonomic units). This represents pollinate monoecious Ficus species have consider- ca 15 percent of all species likely to be found in ably longer ovipositors than those associated with Borneo as a whole, assuming the general 1:1 host: dioecious hosts. Strangler fig trees, with their figs pollinator relationship (Corner 1963, Wiebes placed high in the canopy, are monoecious, whereas 1979). Given that a vertical column of air only 14 dioecious species mainly produce their figs at lower cm in diameter was sampled passively along a frac- levels. Species associated with monoecious hosts tion of its length, for a period that was no longer dominated the catches, both in numbers of indi- than the duration that fig trees would remain at- viduals and species (Fig. 3); 78.6 percent of the tractive to pollinators, this high species richness il- pollinators of the monoecious Ficus were present lustrates the tremendous dispersal ability and host on the highest trap, compared to 64.3 percent of finding potential of the wasps. A small number of those pollinating dioecious hosts (Table 4). This pollinator fig wasp species nonetheless dominated ␹2 ϭ Ͼ difference was not significant ( [1] 1.9, P the total catch, with many of the other species rep- 0.05), suggesting that fig wasps pollinating figs sit- resented by just one or two individuals (Fig. 3). uated at lower elevations may fly as high above the The more abundant taxa may have arisen from a

TABLE 3. Daytime captures of chalcids on sticky traps positioned from ground level (trap 1) to 37 m (trap 5). Period 1 only: ‘‘Other ﬁg wasps’’ were non-pollinators belonging to the subfamilies Epichrysomallinae, Sycoryctinae, and Sycophaginae.

Traps Trap Trap height heights 12345Total (x¯) (range) Fig pollinators 0 1 0 4 2 7 4.0 2–5 Other ﬁg wasps 0 0 0 0 0 0 — — Aphelinidae 0 0 1 0 1 2 4.0 3–5 Chalcididae 1 0 1 0 4 6 4.0 1–5 Encyrtidae 10 11 13 14 22 70 3.4 1–5 Eulophidae 2 0 1 8 5 16 3.9 1–5 Eupelmidae 2 5 5 16 29 57 4.1 1–5 Mymaridae 1 2 0 1 0 4 2.2 1–4 Pteromalidae 2 12 1 4 10 29 3.3 1–5 Total chalcids 18 31 22 47 73 191 3.7 1–5 520 Compton, Ellwood, Davis, and Welch

FIGURE 2. The chalcids (all families, including Agaonidae) captured during daytime (open bars) and nighttime (solid bars) on sticky traps positioned from ground level (trap 1) to 37 m (trap 5) in rain forest at Danum Valley Field Centre, Sabah. small number of fig trees that were releasing fig above), compared to that within and below the wasps in the general area, with the remainder com- canopy, the air column above the general canopy ing from more distant sources; however, no trees offers insects, particularly smaller, slower-flying spe- releasing wasps were found in a thorough search cies, the potential to disperse over much greater within an ca 100-meter radius of the traps, or by distances than they could otherwise achieve. Our a less thorough survey beyond this. hypotheses that fig wasps are more likely to travel above the canopy than other chalcids and that this DISCUSSION would be irrespective of where their host figs were positioned were supported insofar as they were the These results were obtained from a single location, only group that was almost entirely restricted to over a relatively brief period of time. Furthermore, traps placed in the overstory. The opportunities for the traps did not extend above the emergent trees dispersal above the canopy, however, were utilized in the forest. The generality of the results therefore not only by fig wasps, as most other families of must be treated with caution. At Danum Valley at chalcids were also numerous on the traps above the least it is clear that insect activity above the general canopy and were found also at lower levels in the canopy is not insignificant. Because of the much forest. The mymarids (a family of minute egg par- more rapid movement of air in the overstory (and asitoids) appeared to be an interesting exception in that they were restricted largely to the lower levels in the forest. TABLE 4. The trap heights of fig wasps from ground level Physical conditions above the canopy can be (trap 1) to 37 m (trap 5) in relation to the much harsher than below, with higher temperatures pollination system of their host trees (monoecious or dioecious). Four additional individuals and lower humidities. These conditions are likely were too badly damaged for host determina- to pose particular physiological problems for the tion. small insects like chalcids, which might be expected to benefit most from using the ‘‘highway’’ above Traps the forest. We hypothesized that flying at night 1–3 Trap 4 Trap 5 above the canopy, rather than during the day, was Monoecious pollinators 0 12 44 one way in which fig wasps and other chalcids Dioecious pollinators 1 4 9 might circumvent this problem. Relatively more Rain Forest Chalcid Flight Heights 521

Crepuscular activity therefore may be a feature of some of the species. Discussions on vertical stratification of insects in rain forests tend to treat the canopy as the up- permost layer of the forest, ignoring the air above. Yet it clearly cannot be ignored when the dispersal of animals is being considered because the overstory appears to be the preferred region for dispersal of fig wasps, some other chalcids, and perhaps other small insects. Fig trees have a broad importance in many topical forests because their figs provide resources for many species of vertebrates (Terborgh 1986). Forest disturbance, including logging op- erations, not only reduces the densities and spatial distribution of many species of fig trees, but also changes the nature of the physical environment through which fig wasps have to disperse. Knowing how fig wasps get from one tree to another is im- portant if we are to predict how the pollination of FIGURE 3. The frequency of capture for individuals figs, and consequently the availability of figs for belonging to the 21 pollinating fig wasp taxa on sticky frugivores to eat, is being changed by rain forest traps in rain forest at Danum Valley Field Centre, Sabah. exploitation. Open bars: species with dioecious Ficus hosts; solid bars: species with monoecious hosts. ACKNOWLEDGMENTS This work was supported by the Royal Society South East Asia Rain forest Research Programme (Programme Pub- chalcids were trapped above the canopy at night lication No. A/296) and the Canopy Research Pro- gramme of the European Science Foundation. We are than during daylight, but overall chalcid activity grateful to the Economic Planning Unit of the govern- was nonetheless greater during daylight hours. Sev- ment of Malaysia, Yayasan Sabah Forestry Upstream Di- eral pollinator fig wasps with typical diurnal facies vision, Danum Valley Management Committee, Sabah (dark body, small eyes) were recorded on the traps Chief Ministers Department, and Universiti Malaysia Sa- bah for allowing AJD and MDFE to conduct research in that were collected just after dawn. Some were still Sabah. Thanks also to our collaborators in Sabah, asso- alive, indicating that they had been trapped very ciate professor Dr. Maryati Mohamed and Dr. Chey Vun recently and had been flying at or around dawn. Khen.

LITERATURE CITED

BOUCEK, Z. 1988. Australasian Chalcidoidea (Hymenoptera). CAB International, Wallingford, England. 832 pp. COMPTON, S. G. 1993. One way to be a fig. Afr. Entomol. 1: 151–158. CORNER, E. J. H. 1963. Ficus on Mt. Kinabalu. Proc. Linn. Soc. Lond. 175: 37–39. HENSHAW,D.J.DE C. 1993. Specimen recovery from sticky traps. Entomol. Mon. Mag. 129: 236. JANZEN, D. H. 1979. How to be a fig. Annu. Rev. Ecol. Syst. 10: 13–51. KATO, M., T. INOUE,A.A.HAMID,T.NAGAMITSU,M.B.NERDEK,A.R.NONA,T.ITINO,S.YAMANE, AND T. Y UMOTO. 1995. Seasonality and vertical structure of light-attracted insect communities in a dipterocarp forest in Sarawak. Res. Popul. Ecol. 37: 59–79. KIRA,T.,AND K. YODA. 1989. Vertical stratification in microclimate. In H. Leith and M. J. A. Werger (Eds.). Ecosystems of the world 14b. Tropical rain forest ecosystems, pp. 55–72. Elsevier, Amsterdam, The Nether- lands. MAWDSLEY, N., S. G. COMPTON, AND R. J. WHITTAKER. 1998. Population persistence, pollination mutualisms, and figs in fragmented tropical landscapes. Conserv. Biol. 12: 1416–1420. NASON, J. D. 1996. Paternity analysis of the breeding structure of strangler fig populations: evidence for substantial long-distance wasp dispersal. J. Biogeogr. 23: 501–512. NEWBERY D., E. J. F. CAMPBELL,Y.F.LEE,C.E.RIDSDALE, AND M. J. STILL. 1992. Primary lowland dipterocarp forest at Danum Valley, Sabah, Malaysia: structure, relative abundance and family composition. Philosophic. Trans. R. Soc. Lond. B 335: 341–356. 522 Compton, Ellwood, Davis, and Welch

NG, R. 1978. The vertical distribution of aerial insects in Pasoh Forest Reserve. Malay. Nat. J. 30: 299–305. SUTTON, S. L. 1989. The spatial distribution of flying insects. In H. Leith and M. J. A. Werger (Eds.). Ecosystems of the world 14b. Tropical rain forest ecosystems, pp. 427–436. Elsevier, Amsterdam, The Netherlands. , AND P. J. HUDSON. 1980. The vertical distribution of flying insects in the lowland rain forest of Zaire. Zool. J. Linn. Soc. 68: 111–123. TERBORGH, J. 1986. Keystone plant resources in the tropical forest. In M. E. Soule´ (Ed.). Conservation biology: the science of scarcity and diversity, pp. 330–344. Sinauer Associates, Sunderland, Massachusetts. WARE, A. B., AND S. G. COMPTON. 1994a. Fig wasp responses to host plant volatiles. J. Chem. Ecol. 20: 785–802. , AND . 1994b. Dispersal of adult female fig wasps I: arrivals and departures. Entomol. Exp. Appl. 73: 221–230. , AND . 1994c. Dispersal of adult female fig wasps II: movements between trees. Entomol. Exp. Appl. 73: 231–238. ,P.T.KAYE,S.G.COMPTON, AND S. VAN NOORT. 1993. Fig volatiles: their role in attracting pollinators and maintaining pollinator specificity. Plant Syst. Evol. 186: 147–156. WHITMORE, T. C. 1984. Tropical rain forests of the Far East. Clarendon Press, Oxford, England. 282 pp. WIEBES, J. T. 1979. Co-evolution of figs and their insect pollinators. Annu. Rev. Ecol. Syst. 10: 1–12.